Oxygen Reduction Reaction: Rapid Prediction of ... - ACS Publications

Subscriber access provided by STEPHEN F AUSTIN STATE UNIV

Surfaces, Interfaces, and Catalysis; Physical Properties of Nanomaterials and Materials

Oxygen Reduction Reaction: Rapid Prediction of Mass Activity of Nanostructured Platinum Electrocatalysts Marlon Rueck, Aliaksandr S. Bandarenka, Federico Calle-Vallejo, and Alessio Gagliardi J. Phys. Chem. Lett., Just Accepted Manuscript • DOI: 10.1021/acs.jpclett.8b01864 • Publication Date (Web): 20 Jul 2018 Downloaded from http://pubs.acs.org on July 20, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry Letters

Oxygen Reduction Reaction: Rapid Prediction of Mass Activity of Nanostructured Platinum Electrocatalysts

Marlon Rück,

∗,†

Aliaksandr Bandarenka,

‡

¶

Federico Calle-Vallejo,

and Alessio

†

Gagliardi

†Department

of Electrical and Computer Engineering, Technical University Of Munich, 80333 München, Germany

‡Physics

Department, Technical University Of Munich, 85748 Garching, Germany

¶Department

of Materials Science and Physical Chemistry, Institute of Theoretical and

Computational Chemistry (IQTC), University of Barcelona, 08028 Barcelona, Spain

E-mail: [email protected]

1

ACS Paragon Plus Environment

The Journal of Physical Chemistry Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract Tailored Pt nanoparticle catalysts are promising candidates to accelerate the oxygen reduction reaction (ORR) in fuel cells. However, the search for active nanoparticle catalysts is hindered by laborious eort of experimental synthesis and measurements. On the other hand, DFT-based approaches are still time consuming and often not ecient. In this study, we introduce a computational model which enables rapid catalytic activity calculation of unstrained pure Pt nanoparticle electrocatalysts. Regarding particle size eects on Pt nanoparticles, experimental catalytic mass activities from previous studies are accurately reproduced by our computational model. Moreover, beyond available experiments, our computational model identies potential enhancement in mass activity up to 190% over the experimentally detected maximum. Importantly, the rapid activity calculation enabled by our computational model may pave the way for extensive nanoparticle screening to expedite the search for improved electrocatalysts.

Graphical TOC Entry (110)

(111) (111) (110) 2.26 A/mgPt

2


Page 2 of 17

Page 3 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Proton-exchange membrane fuel cells (PEMFCs) are suitable devices for versatile stationary and portable energy solutions. 1 Apart from industrial concerns on the durability of fuel cells, 2 widespread commercialization of fuel cell technologies is impeded by the high costs of platinum which is required for the oxygen reduction reaction (ORR) at the cathode side of these devices. 3 Due to dierent adsorption energies of reaction intermediates, the catalytic activity strongly depends on the catalyst surface structure. 4 Structural sensitivity has been extensively studied on stepped surfaces which harbor considerably increased catalytic activities relative to Pt(111). 57 Nanoparticle catalysts combine high surface to volume ratio with the capability to tailor active catalyst surface structures. Prominently, Pt nanowires have recently been fabricated at laboratory level which exceed the mass activity of current state-of-the-art commercial platinum on carbon supported (Pt/C) catalysts by a factor of 52. 8,9 Nevertheless, the progress in search for promising nanoparticle catalysts is restricted by complex synthesis on the experimental side and approaches solely based on expensive atomistic density functional theory (DFT) on the theory side. 10 Suitable descriptors significantly promote the classication of catalyst structures into promising and inactive. From an early stage, d-band centers 11 have been an important concept in this eld. However, the necessity of DFT calculations for d-band studies further stimulated the identication of more aordable descriptors. To this end, conventional coordination numbers, counting the number of rst nearest neighbors of the active sites, yield appropriate scaling relations for catalytic activities of extended surfaces. 12,13 However, activity trends for nanoparticle catalysts remain out of scope due to nite size eects. 14,15 In recent studies, Calle-Vallejo et al. extended the concept of coordination number to the second nearest neighbors by means of generalized coordination numbers (gCN) ni ∑ cn(j) CN (i) = . cnmax j=1

(1)

Those have been proven simple descriptors for catalytic activities of various reactions. 1619

3



The conventional coordination numbers cn(j) are summed up for all ni nearest neighbor sites j of site i such that nite size eects and local site structures are considered explicitly. The maximal coordination number cnmax (i.e. cnmax = 12 for top sites in the fcc structure) yields a normalization to bulk atoms which are represented by CN = 12 equivalently to the conventional bulk coordination number. The Sabatier analysis for the oxygen reduction reaction, in which competing adsorption energies of the intermediates ∗ OH and ∗ OOH are evaluated similarly to an earlier study, 20 revealed an optimal adsorption energy tradeo in range of 7.5 < CN ≤ 8.3 where enhanced catalytic activity relative to Pt(111) is expected. 17,18 Moreover, the DFT based adsorption energies of all crucial ORR intermediates, namely ∗ O, ∗ O2 , ∗ OH, and ∗ OOH, are linearly related with CN . 16,17 Even though CN is a suitable descriptor to identify active sites, this is the rst time that this concept is used to assess current densities and mass activities of electrocatalysts. On the experimental side, the adsorption potentials of ∗ OH can be obtained at model extended surfaces with respect to Pt(111) by the analysis of cyclic voltammograms. 7,18 Furthermore, catalytic activities can be expressed by kinetic current densities at certain important electrode potentials 6,2123 or using quasi exchange current densities within the Tafel approximation. 5 It is observed that a weakening of ∗ OH adsorption potentials with respect to Pt(111) up to ∼ 0.1 − 0.15V results in larger current densities. 17,18,24,25 Herein, we combine theoretical and experimental data to develop a computational model which calculates the catalytic activity of pure unstrained Pt nanoparticle electrocatalysts in a short computation time. We calculate ∗ OH adsorption energies with reference to OH in the gas phase on diversely coordinated sites in multifaceted nanocatalyst shapes (as tetrahedrons, cuboctahedrons, truncated octahedrons and extended surfaces) by means of DFT. Note, however, that the dierence between two adsorption energies does not depend on the gas-phase reference used, as long as the reference is identical. For details on the DFT calculations we refer the reader to our Supporting Information.

4


Page 4 of 17

Page 5 of 17

b)

a)

(110)

(111)

(111) 2 nm

(110) 1 nm

(110)

d = 2.86 nm

(110) (111)

(111)

) 11

(1

(111)

(111)

(110)

(110) 3.3 nm 10

Experimental activity at 0.9 V vs RHE relative to Pt(111)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


1

c)

~1/2 ML n=9

Pt(111)

0.000

0.025

0.050

d)

0.075

0.100

A2

~2/3 ML Pt3Ni(111) Pt3ML/Pd(111)

n=2

n=2

Pt[n(111)x(111)] Pt[n(111)x(100)] Pt3Ni[n(111)x(111)] Pt3Ni[n(111)x(100)] Pt3Co[n(111)x(111)] Pt3Co[n(111)x(100)]

EOH

6.2 nm

n=9

n=4

Pt/Pd(111)

(110)

n=3

__ n=3 n=4 n=9 n=5 n=2 __ n=3 n=9 n=9 n=3 n=3 Pt3Co(111) n=2 n=3 ~1/3 ML n=5 __ n=7 n=7 n=6 Cu/Pt(111) NSA

n=9

Pt3Ni(111) NSA

1)

1 (1

0.125

Pt/Pd/Pd3Fe(111)

A1

n=2

MAE = 1.1, MAX = 4.57

n=2

A1(CN) = exp[3.14CN

23.40]

A2(CN) = exp[ 4.96CN + 42.18]

n=2

0.150

0.175 7.4

Peak (A1(CNp) = A2(CNp)) at CNp = 8.10

7.6

Pt(111) EOH /V

7.8

8.0

8.2

8.4

8.6

CN

Figure 1: (a) Adsorption energies ∆EOH are calculated by DFT (see Supporting Information) on diversely coordinated sites in various catalyst shapes of dierent sizes; including tetrahedrons (green), cuboctahedrons (magenta), truncated octahedrons (yellow, cyan, grey, brown), extended surfaces (orange) and cavities (blue). The linear dependence on CN is described by the linear function provided in the inset. The gray area covering ±2MAE around the linear t is provided that contains 75% of the calculated data points. (b) Spherical nanoparticles (printed by ASE 26 ) at diameters of 1 nm, 2 nm, 3.3 nm and 6.2 nm are exemplied as they are investigated in this study. Active sites with 7.5 ≤ CN ≤ 8.3 are highlighted in yellow. Prominent low index surfaces are enclosed by dashed lines. (c) Relative experimental activities of various Pt stepped surfaces (forming terrace widths of length n) and Pt alloy fcc(111) single-crystals are plotted vs experimental ∗ OH binding energies: (black open squares) Pt stepped surfaces; (green open squares) Pt3 Ni stepped surfaces; (red open squares) Pt3 Co stepped surfaces; (red up-pointing traingle) Cu/Pt(111) NSAs with full and partial (1/3 ML, 1/2 ML, 2/3 ML) surface Cu content; (full green down-pointing triangle) Pt3 Ni(111) NSA; (open green down-pointing triangle) bulk Pt3 Ni(111); (blue star) one monolayer of Pt on Pd(111); (blue plus) monolayer of Pt on annealed Pd3 Fe(111) electrode with one segregated Pd layer; (open blue octahedron) three monolayers of Pt on Pd(111); (blue x) bulk Pt3 Co(111). Pt stepped surfaces, which are highlighted by underlined terrace widths of length n = 3, n = 4, n = 7, are taken from Ref. 18 Remaining data is taken from Ref. 7 and sources therein. The catalytic activities are measured at 0.9 V vs RHE in 0.1 M HCLO4 . (d) The linear scaling relation in a) maps the experimental binding energy in c) onto CN . Linear regression data of the increasing and decreasing activity functions A1 and A2 , respectively, is provided in the inset. The gray area covering ±2MAE around the linear t contains ∼ 92% of the calculated data points. 5 ACS Paragon Plus Environment


The evaluation of coordination for all involved sites exposes fundamental linear dependence between ∗ OH adsorption energies and CN , which we present in Figure 1a, and the associated error analysis is included in the Supporting Information. In addition, we use experimental data from literature, 7,18 comprising catalytic activities versus experimentally observed ∗ OH binding energies for extended Pt surfaces and Pt alloys, to draw the volcano plot in Figure 1c. Importantly, the above discussed linear relation from Figure 1a is then employed to map the experimental ∗ OH binding energies from Figure 1c to an equivalent of

CN . The procedure is outlined in the Supporting Information. As shown in Figure 1c, the Pt alloys form an activity peak at ∗ OH binding potential of

∼ 0.13 V which involves activity enhancement of 800% relative to Pt(111). The lower activity peak for Pt stepped surfaces is located at weaker ∗ OH binding potential of ∼ 0.1 V . However, it becomes apparent that Pt alloys and pure Pt surfaces follow the same activity trend. For a given surface facet, it is known that dierences in ∗ OH adsorption energies between Pt stepped surfaces and Pt alloys are governed by strain and/or ligand eects. For the Pt alloys in Figure 1c, it is argued in Ref. 7 that strain eects predominate over ligand eects, because ligand eects decay very fast as a function of interatomic distances. Strain eects are rather assumed to be important. Recently, it has been shown that the concept of gCN can be extended to strain eects by means of the generalized coordination number ∗

CN . 27 Strained catalysts follow the same trends as the unstrained catalysts in terms of ∗

CN and CN , respectively. Hence, one should notice that there is no discrepancy between the fact that Pt alloys and pure Pt surfaces are used to construct the resulting volcano plot in Figure 1d which is employed in our computational model. The associated volcanoshaped catalytic activity trend agrees well with the aforementioned Sabatier analysis 17 where enhanced catalytic activities relative to Pt(111) are expected for sites with generalized coordination 7.5 < CN ≤ 8.3. The trend is captured by t functions A1 and A2 (see Supporting Information) which form the peak of the volcano at CN = 8.1. This corresponds to an ∗ OH binding potential relative to Pt(111) of ∆EOH − ∆E P t(111) ≈ 0.115 V . Thus, unstrained 6


Page 6 of 17

Page 7 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Pt nanoparticle catalysts can be examined by evaluation of gCNs at all nanoparticle sites. To this end, the activity contributions of all sites are summed up according to the trend in Figure 1d. We discuss this essential step in more detail in the Supporting Information where an error analysis is taken into account. Even more intriguingly, we exploit additional geometrical considerations and the total number of sites (which is pointed out in the Supporting Information) in order to yield mass activities not only relative to Pt(111), but rather in units of Amperes per milligram of Pt. Beyond the peak of the volcano at larger CN in Figure 1d, the activity trend is widely dispersed. At small CN , undercoordinated sites may be aected by oxygenated species which leads to blocked catalytic processes at these centers. 28 Therefore, the activity contribution of sites with CN < 7.5 or CN > 8.3 is set to zero in our computational model, but all nanoparticle sites are taken into account for the mass activity prediction. It is noteworthy that our computational model does not employ any additional assumptions than those general considerations discussed above. ∗

Moreover, using the above-discussed generalized coordination number CN to describe strain eects, 27 our computational model has been straightforwardly extended to strained nanoparticle electrocatalysts which is presented on an exemplied degradation study in the SI. The ∗

∗

descriptors CN and CN dier in nearest neighbor weighting. In CN the nearest neighbors are weighted with the ratio of interatomic distance per interatomic bulk distance. Thus, ∗

using the more general descriptor CN , instead of CN , to comprise strain eects does essentially not lead to increased computation times in our approach. In this study, the nanoparticle catalysts are modeled by quasi-spherical shapes such as those exemplied in Figure 1b. Additional spherical nanoparticle catalysts comprising a broader range of diameters are presented in Figure S1 in the Supporting Information. Within our computational model, the mass activities for 620 distinct nanoparticles are evaluated for diameters ranging from 0.6 nm to 13 nm in small-scale 0.02 nm intervals. As it becomes apparent in Figure 2, the mass activity depends sensitively on the nanoparticle diameter. This structure sensivity becomes increasingly pronounced towards smaller diameters where 7



local site structures are strongly aected upon slight changes in the diameter which is in line with the study in Ref. 29 This insight is further substantiated by the relationship between nanoparticle sizes and the number of active sites with 7.5 ≤ CN ≤ 8.3 which is presented in the Supporting Information. The overall mass activity trend features a peak near 2.5 nm. Thus, the analysis of Figure 2 also shows that the size distribution of the nanoparticles turns out to be crucial for accurate catalytic activity prediction. Therefore, activities for distinct nanoparticle diameters are obtained by the mean activity within the diameter distribution.

Mass activity at 0.9 V vs RHE (A/mg Pt)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 17

2.5

Computational model Mass activity trend

2.0

1.5

1.0

Structure sensitivity

0.5

0.0 0

2

4

6

8

10

Nanoparticle diameter (nm)

12

Figure 2: Computational results for the mass activity, which is obtained in absolut units of A/mgP t , versus nanoparticle sizes. 620 distinct nanoparticles, involving diameters between 0.6 nm and 13 nm in intervals of 0.02 nm, are taken into account. The dashed red curve depicts the overall mass activity trend which features a peak near 2.5 nm. Structure sensitivity, shown by the vertical bar, indicates that strong coordination is abundant at the top of the bar, whereas low coordination is abundant at the bottom of the bar. The applicability of the computational model is further compared with experimental data. Perez-Alonso et al.

and Shao et al. independently investigated nanoparticle size eects on

the catalytic activity as shown in Figure 3a and Figure 3b, respectively. 22,23 Although the experimental nanostructures were not characterized in all details on the atomic scale, the results from electron microscopy in both experiments give reason to model the nanoparticles by spherical shapes as we did in this study. Note that the maximal mass activity has been detected at nanoparticle sizes between 2-3 nm which coincides with related experimental 3,30 8


a)

Experiment DFT Computational model

1.0

0.8

0.6

0.4

0.2

2

4

6

8

10


12


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60



Page 9 of 17

14

0.12

b)

Experiment Computational model (×0.09)

0.10

0.08

0.06

0.04

1

2

3

4


5

Figure 3: Experimental (red dots) and computational results (green dots) of the particle size eect. Vertical error bars for computational mass activities represent the standard error of the mean. (a) Experimental data is taken from Perez-Alonso et al. 23 Computational mass activities are calculated and displayed in absolute units of A/mgP t . Diameter distributions employed in the computational model are adopted from the experimental study. (b) The experimental activities in this particular study from Shao et al. 22 dier from a) by one magnitude. Thus, unlike the absolute approach in a), all computational mass activities are multiplied by a factor of 0.09 as a t to the experimental data. The standard deviation of the diameter is constrained between 0.18-0.35 nm similar to the experimental specication of 0.2-0.3 nm. and theoretical 31,32 studies. For the dataset in Figure 3a, the experimental diameter distribution is specied individually for each nanoparticle. We equally adapt the experimental diameter distributions in our computational model. Interestingly, the experimental mass activity trend is precisely reproduced by our computational approach. Furthermore, particularly regarding absolute units the computational and experimental mass activities coincide as the associated error intervals overlap; except for the smallest diameter near 2 nm where corrosion eects are believed to have degraded the experimental nanoparticle structure. 23 In this regard, it is important to emphasize that slight deviations in the size distribution may considerably aect the associated mass activity around diameters of 2 nm. Note that for computational studies which are focussed on eects from nanoparticle degradation on the mass activity, the aforementioned extension of our computational model using CN

∗ 27

is applicable. The second experimental dataset in Figure 3b comprises signicantly lower mass activities at a level of 10% compared to the mass activities in Figure 3a. Consequently, 9



unlike the absolute approach in Figure 3a, the computational activity trend for the second experimental dataset in Figure 3b is scaled to t the corresponding experimental trend. Multiplying all computational values by a scale factor of 0.09 yields the best agreement with these particular experiments. Within the computational model, the standard deviation of the diameter is constrained between 0.18 nm and 0.35 nm. For the experimental measurements, the overall standard deviation is stated to be similarly between 0.2 nm and 0.3 nm. 22 As remarkable result, the computational trend is in good agreement with the experimental values. Furthermore, it is important to note that the steep decrease in experimental activity around 2 nm in Figure 3a is considerably less pronounced in Figure 3b. This substantiates the assumption that corrosion has aected the surface structure of the 2 nm nanoparticles for the case shown in Figure 3a. Perez-Alonso et al.

compared the experimental results with an earlier theoretical study 25,31

which is represented by the dashed curve in Figure 3a. Therein, nanoparticles are constructed by edged surface facets which dier from spherical shapes. Relative activities in arbitrary units are obtained via adsorption free energies from DFT calculations. The experimental trend is adequately captured in the sense that the mass activity peak at 2-4 nm is reproduced which is followed by a slightly attened decrease in mass activity towards large diameters. However, the precision in nanoparticle size has not been taken into account in this DFT approach. Consequently, experimental and theoretical approaches still need to be brought in quantitative agreement. Remarkably, this has been achieved in the present computational model by explicit consideration of size distribution and absolute units, which constitutes a step forward compared to previous studies. Furthermore, such quantitative agreement with experimental data ascertains that spherical nanoparticles serve as the appropriate model structures in order to simulate real nanoparticle catalysts. The activity analysis of our computational model enables interesting nanoparticle size pre10


Page 10 of 17

Page 11 of 17

dictions with enhanced activity performance. Exploring the nanoparticle size eect at the maximum level of detail, we produced the contour plot in Figure 4a where the nanoparticle diameter range and associated diameter distributions are mapped onto the catalytic activity. The experimental dataset from Figure 3a is shown in this contour plot by black dots. The contour plot unveils the highest potential for mass activity improvement at nanoparticle diameters of 1 nm, 2 nm and 2.9 nm for nanoparticle size distributions below 0.2 nm. Those nanoparticles harbor mass activity enhancement of 152%, 178% and 190% at (1.0±0.1) nm, (2.0±0.1) nm and (2.9±0.1) nm, respectively, compared to the highest experimental mass activity in Figure 3a. Recently realized elaborate fabrication methods enable such precise size control of Pt nanoparticle catalysts even down to the subnanometer scale 33 giving rise to large catalytic activities at (0.9±0.1) nm nanoparticle size. This result corresponds perfectly

Nanoparticle diameter distribution (nm)

with the computationally predicted activity peak at (1.0 ± 0.1) nm in Figure 4.

1.9

1.91

1.7

1.69

1.5

1.48

1.3

1.27

1.1

1.06

0.9

0.85

0.7

0.64

0.5

0.42

0.3

0.21

0.1

1

2

3

4

5

6


7

8

0.00


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 4: The contour plot elucidates the full parameter space of the particle size eect: Nanoparticle diameters (on the horizontal axis) and associated diameter distributions (on the vertical axis) are mapped onto the catalytic mass activity (presented by the color bar) in absolute units of A/mgP t . The experimental data from Figure 3a (labeled by black dots) is included. The contour plot reveals that highest mass activities (indicated by red colored areas) are harbored by nanoparticles at diameters of 1 nm, 2 nm and 2.9 nm with diameter distributions below 0.2 nm. To conclude, we have presented a computational model which enables rapid activity cal11



Page 12 of 17

culation of 3D Pt unstrained nanoparticle catalysts. In line with experiments, DFT studies show a linear scaling relation between ∗ OH adsorption energies and generalized coordination numbers for Pt. We capitalize here on this crucial result to provide a link between the generalized coordination numbers and experimentally measured ORR catalytic activities. Making use of fundamental geometrical considerations, the presented computational model comprises the capability to determine nanoparticle mass activities in absolute units of A/mg , without the need for a reference to e.g. Pt(111). In this way, expensive DFT calculations are omitted during runtime realizing sharply reduced and feasible computation times in comparison to theoretical approaches which are based exclusively on DFT. The applicability of our computational model was tested on two experimental datasets involving particle size effects. Remarkably, the computational model accurately reproduces the experimental trends. Regarding the absolute units, the mass activities in both experiments dier considerably by one order of magnitude. Nonetheless, quantitative agreement in absolute units has been precisely observed between the computational model and one experimental dataset. Besides the capability to capture experimental activities on a highly accurate level, this study gives rise to predictions beyond currently available experiments. Promising nanoparticles, which harbor high mass activities, are predicted for nanoparticles sizes near 1 nm, 2 nm and 3 nm with size distributions below 0.2 nm. It is important to note that this complete nanoparticle size eect study was carried out within only few hours by means of the presented computational model. Thus, we believe that rapid nanoparticle activity calculation paves the way for high-throughput nanoparticle activity screening, which may strongly expedite the search for innovative catalysts in future studies.

Acknowledgement This work is supported by the German Research Foundation (DFG) under Grant No. 355784621 and the excellence cluster Nanosystems Initiative Munich (NIM) of the DFG. FCV thanks

12


Page 13 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Spanish MEC for a Ramón y Cajal research contract (RYC-2015-18996).

Supporting Information Available The Supporting Information includes details of the DFT cal-

Supporting Information.

culations, shapes of spherical nanoparticle catalysts, details of the computational model, an analysis of the number of active sites, a showcase for the prediction of catalytic mass activities, and an exemplied degradation study using our computational model extended to strain eects. This material is available free of charge via the Internet at http://pubs.acs.org/.

References (1) Sharaf, O. Z.; Orhan, M. F. An Overview of Fuel Cell Technology: Fundamentals and Applications. Renew. Sustain. Energy Rev.

2014,

32,

810853.

(2) Wu, J.; Yuan, X. Z.; Martin, J. J.; Wang, H.; Zhang, J.; Shen, J.; Wu, S.; Merida, W. A Review of PEM Fuel Cell Durability: Degradation Mechanisms and Mitigation Strategies. J. Power Sources

2008,

184,

104119.

(3) Gasteiger, H. A.; Kocha, S. S.; Sompalli, B.; Wagner, F. T. Activity Benchmarks and Requirements for Pt, Pt-Alloy, and Non-Pt Oxygen Reduction Catalysts for PEMFCs. Appl. Catal. B Environ.

2005,

56,

935.

(4) Duan, Z.; Wang, G. A First Principles Study of Oxygen Reduction Reaction on a Pt(111) Surface Modied by a Subsurface Transition Metal M (M = Ni, Co, or Fe). Phys. Chem. Chem. Phys.

2011,

13,

2017820187.

(5) Macia, M. D.; Campina, J. M.; Herrero, E.; Feliu, J. M. On the Kinetics of Oxygen

13



Page 14 of 17

Reduction on Platinum Stepped Surfaces in Acidic Media. J. Electroanal. Chem. 2004, 564,

141150.

(6) Hitotsuyanagi, A.; Nakamura, M.; Hoshi, N. Structural Eects on the Activity for the Oxygen Reduction Reaction on n(111)-(100) Series of Pt: Correlation with the Oxide Film Formation. Electrochim. Acta

2012,

82,

512516.

(7) oli¢, V.; Bandarenka, A. S. Pt Alloy Electrocatalysts for the Oxygen Reduction Reaction: From Model Surfaces to Nanostructured Systems. ACS Catal. 2016, 6, 53785385. (8) Li, M.; Zhao, Z.; Cheng, T.; Fortunelli, A.; Chen, C. Y.; Yu, R.; Zhang, Q.; Gu, L.; Merinov, B. V.; Lin, Z. et al. Ultrane Jagged Platinum Nanowires Enable Ultrahigh Mass Activity for the Oxygen Reduction Reaction. Science

2016,

354,

14141419.

(9) Stephens, I. E. L.; Rossmeisl, J.; Chorkendor, I. Toward Sustainable Fuel Cells Improved. Science

2016,

354,

13781379.

(10) Viñes, F.; Gomes, J. R. B.; Illas, F. Understanding the Reactivity of Metallic Nanoparticles: Beyond the Extended Surface Model for Catalysis. Chem. Soc. Rev.

2014,

43,

49224939. (11) Hammer, B.; Nørskov, J. K. Theoretical Surface Science and CatalysisCalculations and Concepts. Adv. Catal.

2000,

45,

71129.

(12) Li, H.; Li, Y.; Koper, M. T.; Calle-Vallejo, F. Bond-Making and Breaking Between Carbon, Nitrogen, and Oxygen in Electrocatalysis. J. Am. Chem. Soc.

2014,

136,

1569415701. (13) Calle-Vallejo, F.; Loreda, D.; Koper, M. T. M.; Sautet, P. Introducing Structural Sensitivity into Adsorption-Energy Scaling Relations by means of Coordination Numbers. Nat. Chem.

2015,

7,

403410.

14


Page 15 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(14) Kleis, J.; Greeley, J.; Romero, N. A.; Morozov, V. A.; Falsig, H.; Larsen, A. H.; Lu, J.; Mortensen, J. J.; Duªak, M.; Thygesen, K. S. et al. Finite Size Eects in Chemical Bonding: From Small Clusters to Solids. Catal. Letters

2011,

141,

10671071.

(15) Mpourmpakis, G.; Andriotis, A. N.; Vlachos, D. G. Identication of Descriptors for the CO Interaction With Metal Nanoparticles. Nano Lett.

2010,

10,

10411045.

(16) Calle-Vallejo, F.; Martínez, J. I.; García-Lastra, J. M.; Sautet, P.; Loreda, D. Fast Prediction of Adsorption Properties for Platinum Nanocatalysts with Generalized Coordination Numbers. Angew. Chemie - Int. Ed.

2014,

53,

83168319.

(17) Calle-Vallejo, F.; Tymoczko, J.; Colic, V.; Vu, Q. H.; Pohl, M. D.; Morgenstern, K.; Loreda, D.; Sautet, P.; Schuhmann, W.; Bandarenka, A. S. Finding Optimal Surface Sites on Heterogeneous Catalysts by Counting Nearest Neighbors. Science

2015,

350,

185189. (18) Calle-Vallejo, F.; Pohl, M. D.; Reinisch, D.; Loreda, D.; Sautet, P.; Bandarenka, A. S. Why Conclusions from Platinum Model Surfaces Do Not Necessarily Lead to Enhanced Nanoparticle Catalysts for the Oxygen Reduction Reaction. Chem. Sci. 2017, 8, 2283 2289. (19) Strasser, P.; Gliech, M.; Kuehl, S.; Moeller, T. Electrochemical Processes on Solid Shaped Nanoparticles with Dened Facets. Chem. Soc. Rev.

2018,

47,

715735.

(20) Nørskov, J. K.; Rossmeisl, J.; Logadottir, A.; Lindqvist, L.; Kitchin, J. R.; Bligaard, T.; Jónsson, H. Origin of the Overpotential for Oxygen Reduction at a Fuel-Cell Cathode. J. Phys. Chem. B

2004,

108,

1788617892.

(21) Stamenkovic, V. R.; Fowler, B.; Mun, B. S.; Wang, G.; Ross, P. N.; Lucas, C. A.; Markovic, N. M. Improved Oxygen Reduction Activity on Pt3Ni(111) via Increased Surface Site Availability. Science

2007,

315,

15

493497.



Page 16 of 17

(22) Shao, M.; Peles, A.; Shoemaker, K. Electrocatalysis on Platinum Nanoparticles: Particle Size Eect on Oxygen Reduction Reaction Activity. Nano Lett. 2011, 11, 37143719. (23) Perez-Alonso, F. J.; McCarthy, D. N.; Nierho, A.; Hernandez-Fernandez, P.; Strebel, C.; Stephens, I. E. L.; Nielsen, J. H.; Chorkendor, I. The Eect of Size on the Oxygen Electroreduction Activity of Mass-Selected Platinum Nanoparticles. Angew. Chemie - Int. Ed.

2012,

51,

46414643.

(24) Bandarenka, A. S.; Hansen, H. A.; Rossmeisl, J.; Stephens, I. E. L. Elucidating the Activity of Stepped Pt Single Crystals for Oxygen Reduction. Phys. Chem. Chem. Phys.

2014,

16,

1362513629.

(25) Stephens, I. E. L.; Bondarenko, A. S.; Grønbjerg, U.; Rossmeisl, J.; Chorkendor, I. Understanding the Electrocatalysis of Oxygen Reduction on Platinum and its Alloys. Energy Environ. Sci.

2012,

5,

67446762.

(26) Larsen, A.; Mortensen, J.; Blomqvist, J.; Castelli, I.; Christensen, R.; Dulak, M.; Friis, J.; Groves, M.; Hammer, B.; Hargus, C. et al. The Atomic Simulation Environment - A Python Library for Working with Atoms. J. Phys. Condens. Matter 29,

2017,

273002273031.

(27) Calle-Vallejo, F.; Bandarenka, A. Enabling Generalized Coordination Numbers to Describe Strain Eects. ChemSusChem

2018,

11,

18241828.

(28) Pohl, M. D.; Colic, V.; Scieszka, D.; Bandarenka, A. S. Elucidation of Adsorption Processes at the Surface of Pt(331) Model Electrocatalysts in Acidic Aqueous Media. Phys. Chem. Chem. Phys.

2016,

18,

1079210799.

(29) Sanchez, S. I.; Menard, L. D.; Bram, A.; Kang, J. H.; Small, M. W.; Nuzzo, R. G.; Frenkel, A. I. The Emergence of Nonbulk Properties in Supported Metal Clusters: Negative Thermal Expansion and Atomic Disorder in Pt Nanoclusters Supported on gamma-Al2O3. Journal of the American Chemical Society 16


2009,

131,

70407054.

Page 17 of 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(30) Leontyev, I. N.; Belenov, S. V.; Guterman, V. E.; Haghi-Ashtiani, P.; Shaganov, A. P.; Dkhil, B. Catalytic Activity of Carbon-Supported Pt Nanoelectrocatalysts. Why Reducing the Size of Pt Nanoparticles is not Always Benecial. J. Phys. Chem. C 115,

2011,

54295434.

(31) Tritsaris, G. A.; Greeley, J.; Rossmeisl, J.; Nørskov, J. K. Atomic-Scale Modeling of Particle Size Eects for the Oxygen Reduction Reaction on Pt. Catal. Letters 141,

2011,

909913.

(32) Tripkovi¢, V.; Cerri, I.; Bligaard, T.; Rossmeisl, J. The Inuence of Particle Shape and Size on the Activity of Platinum Nanoparticles for Oxygen Reduction Reaction: A Density Functional Theory Study. Catal. Letters

2014,

144,

380388.

(33) Yamamoto, K.; Imaoka, T.; Chun, W. J.; Enoki, O.; Katoh, H.; Takenaga, M.; Sonoi, A. Size-Specic Catalytic Activity of Platinum Clusters Enhances Oxygen Reduction Reactions. Nat. Chem.

2009,

1,

397402.

17


Oxygen Reduction Reaction: Rapid Prediction of ... - ACS Publications

Recommend Documents